Intel Skylake / Kaby Lake

[Sweepr]

Found this comparison between Skylake with/without eDRAM @ Cinebench R15.

i7-6700HQ and i7-6770HQ have the exact same clockspeeds and TDP (2.6-3.5 GHz, 45W), so it should be reasonably fair. 128MB eDRAM improves performance by 8-9% here:

Will LGA Skylake-S GT4e make it to retail? Could be an interesting part, while we wait for 2018 Coffee Lake (mainstream 6C/12T + GT3 with eDRAM).


Also related, AnandTech has an article about memory scaling on Skull Canyon:

I suspect that the processor configuration (128MB eDRAM) is playing a major role in ensuring that the experience with normal real-world workloads remains the same irrespective of whether it is running a 2400 MHz C16 kit or a 3000 MHz C16 one. The eDRAM keeps the obvious memory bottleneck - the iGPU - from being bandwidth starved, and it can also better feed desktop applications in some situations as well. However if that's the case, then desktop and notebook system users with a different non-eDRAM processor might observe different results.

http://www.anandtech.com/show/10602/memory-frequency-scaling-on-skull-canyon

 

[jpiniero]

I don't think Skylake-C is going to get released.

 

[Sweepr]

DigiTimes spills the beans about Kaby Lake and Intel's future CPUs:

Intel begins shipping Kaby Lake notebook CPUs

Intel has begun delivering its next-generation notebook-use Kaby Lake processors and will start shipping comparable desktop-use CPUs by year-end 2016, according to sources from Taiwan PC makers.

Due to the availability of the notebook Kaby Lake CPUs, notebook vendors have lowered prices for existing Sky Lake-based models, said the sources, noting that prices of some gaming notebooks build on Sky Lake CPUs and Nvidia's last-generation GPUs have been cut by 10%.

Volume shipments of desktop-use Kaby Lake CPUs will come at the end of 2016 and new motherboards based on Z270 and H270 chips are expected to hit vendors' shelves by year-end, indicated the sources.

To pave the way for Kaby Lake desktop processors, Intel has also adjusted the price structure of its 100-series chipsets, encouraging motherboard makers to purchase more H110 chips, said the sources.

According to Intel's roadmap, the chipmaker is expected to roll out its 10nm notebook-use Cannon Lake and 14nm desktop-use Coffee Lake CPUs in the second half of 2017 and to bring out Ice Lake CPUs in 2018.

www.digitimes.com/news/a20160904PD201.html

 

[mikk]

Sounds a bit like that after Coffe Lake Intel will jump directly to Icelake for Desktop. Maybe Coffee Lake is really a Cannonlake in 14+ nm, in this case it would make sense to skip Cannonlake.

 

[Sweepr]

Long awaited Macbook Pro 2016 should be announced tomorrow. I wonder if Apple going for Kaby Lake or Skylake with Iris Graphics in this round. In other news:

Here's Why Intel Kaby Lake And AMD Zen Will Only Be Optimized On Windows 10

There are some additional wrinkles to consider with Kaby Lake that Zen that no one seems to be talking about, however. Both of these microarchitectures have new features and technologies that require significant updates to Windows 10 to optimally function. Kaby Lake, for example, has updates to Intel’s Speed Shift technology that make it possible to change power states more quickly than Skylake. Because Kaby Lake is able to make Speed Shift transitions measurably faster, 7th Gen Core processors based on the architecture will be able to ramp up (and spool down) clocks in as quickly as 15ms. While Speed Shift is enabled in hardware, it requires OS-level hooks to function properly. In Intel’s own presentations in the lead-up to Kaby Lake’s unveiling, the company mentioned that Windows 10 Redstone 1 was used to evaluate the latest updates to Speed Shift in their performance data.

Another thing to consider is Intel’s Turbo Boost Technology 3.0, which was introduced with Broadwell-E. With Turbo Boost Max 3.0 technology, single-thread performance can be improved by identifying the fastest core on a particular processor die and prioritizing critical workloads for that core. With Turbo Boost 3.0, not only does the processor’s frequency ramp up when needed, but workloads are also directed to the fastest possible core available. As it stands today, if you don't have Redstone 1, Turbo Boost Max 3.0 technology requires the installation of a driver and secondary utility that launches with Windows. We assume some form of Turbo Boost 3.0 will be implemented on Kaby Lake-based 7th Gen Core processors for desktop systems as well, and ideally, support for that technology would be baked right into the operating system.

http://hothardware.com/news/microsoft-to-optimize-windows-10-for-kaby-lake-and-zen

Sounds a bit like that after Coffe Lake Intel will jump directly to Icelake for Desktop. Maybe Coffee Lake is really a Cannonlake in 14+ nm, in this case it would make sense to skip Cannonlake.

That's a very agressive approach from Intel. Desktop Kaby Lake in a few months, 6C/12T LGA 1151 Coffee Lake next year, followed by a new CPU architecture (Icelake) in 2018.

If Coffee Lake was a late addition to the game I expect another round of Skylake cores + Gen 9/9.5 iGPU.

 

[Sweepr]

Kaby Lake-U i7-7500U (15W) vs Bristol Ridge FX-9800P (15W)

Cinebench R11.5 - CPU Single 64Bit
FX-9800P: 0.91
i7-7500U: 1.68

Cinebench R11.5 - CPU Multi 64Bit
FX-9800P: 2.86
i7-7500U: 4.03

Cinebench R15 - CPU Single 64Bit
FX-9800P: 74
i7-7500U: 145

Cinebench R15 - CPU Multi 64Bit
FX-9800P: 236
i7-7500U: 363

Geekbench 3 - 64 Bit Single-Core Score
FX-9800P: 2057
i7-7500U: 3706

Geekbench 3 - 64 Bit Multi-Core Score
FX-9800P: 5496
i7-7500U: 7835

3DMark 11 - Performance 1280x720
FX-9800P: 1794
i7-7500U: 1864

3DMark 11 - Performance GPU 1280x720
FX-9800P: 1852
i7-7500U: 1668

http://www.notebookcheck.com/Kaby-Lake-Core-i7-7500U-im-Test-Skylake-auf-Steroiden.172422.0.html
http://www.notebookcheck.com/AMD-FX-9800P-APU-Bristol-Ridge-Benchmarks-und-Specs.173391.0.html

Comparing the performance of Intel and AMD's fastest 15W processors for 2016.

 

[Abwx]

Kaby Lake-U i7-7500U (15W)


http://www.notebookcheck.com/Kaby-Lake-Core-i7-7500U-im-Test-Skylake-auf-Steroiden.172422.0.html
http://www.notebookcheck.com/AMD-FX-9800P-APU-Bristol-Ridge-Benchmarks-und-Specs.173391.0.html

Comparing the performance of Intel and AMD's fastest 15W processors for 2016.

15W KBL you said ?..

Here the one you linked at NBC :

PL2 at 25W, so benched at 25W and stated as being scores for 15W....

 

[Sweepr]

15W KBL you said ?..

Here the one you linked at NBC :

PL2 at 25W, so benched at 25W and stated as being scores for 15W....

After a period of maxing out PL2, it must go below PL1 for a while to stay within the PL1 envelope, but of course you know this. Here's a hint from the pic you used as 'proof', but actually shows the opposite: CPU Package Power = 14.898W

Until you provide data that supports the idea that Kaby Lake-U can sustain a 25W PL2 state in all benchmarks (and during their entire run), I will take these comments as trolling.

 

[LTC8K6]

Not sure how anyone can look at that pic and claim it was benched at 25W.

Even the max power is only 20W.

PL2 is short term burst, and PL1 is sustained, so even PL2 apparently only went to 20W briefly at some point.

At any rate, why is AMD not screaming about the wattage cheating, Abwx? Surely they can see what you can see, and would at least point it out?

 

[Abwx]

After a period of maxing out PL2, it must go below PL1 for a while to stay within the PL1 envelope, but of course you know this

Yes, and i also know that the delay is 2 minutes (or is it 200s ?), about enough to do some shorts benches like CB at full power...

 

[Sweepr]

Yes, and i also know that the delay is 2 minutes (or is it 200s ?), about enough to do some shorts benches like CB at full power...

Apparently you didn't even bother to read i7-7500U's review at all - the data is there. Once again, you are completely wrong:

 

[Abwx]

Not sure how anyone can look at that pic and claim it was benched at 25W.

Even the max power is only 20W.

PL2 is short term burst, and PL1 is sustained, so even PL2 apparently only went to 20W briefly at some point.

It is not briefly since on H265 video play the CPU rate is about 80% for 16.753W only for the cores, uncore is not even counted, so that s 20W permanently on a laptop that is hot.

If it was briefly the thing would stutter, wich is apparently not the case according to NBC...

 

[Abwx]

Apparently you didn't even bother to read i7-7500U's review at all - the data is there. Once again, you are completely wrong:

I saw all this, and this is usefull but not the way you think about it..

Since HVInfo report CPU powers that are of the same level as what they display from being drained at the main this imply that the laptop can be fed from both main and batteries simultaneously, this render the measurements at the main not reliable but this has the advantage for OEMs of cheaper power supplies..

 

[Sweepr]

I saw all this, and this is usefull but not the way you think about it..

Since HVInfo report CPU powers that are of the same level as what they display from being drained at the main this imply that the laptop can be fed from both main and batteries simultaneously, this render the measurements at the main not reliable but this has the advantage for OEMs of cheaper power supplies..

Nope, the only thing this implies is that your conspiracy theories are way out of touch with reality. CB11.5 MT power started at 18.2W and a few seconds later was already down to 14.9W (TDP), a far cry from the imaginary 'PL2 25W state for 2 minutes'. i7-7500U sustains same/higher clocks than i7-6600U while using almost 5W less, which is huge for mobile.

 

[Sweepr]

Taking a second look at Bristol Ridge scores here + AMD's presentations.

AMD results:

NotebookCheck results:

- Cinebench R15 - CPU Single 64Bit
FX-9800P: 74
i7-7500U: 145
- Cinebench R15 - CPU Multi 64Bit
FX-9800P: 236
i7-7500U: 363

Other than the fact that i7-7500W (15W) is still considerably faster than their 35W offering - how can FX-9800P beat the performance of a higher-power part?

 

[Abwx]

Taking a second look at Bristol Ridge scores here + AMD's presentations:

For one that s not the AMD presentation but viral marketing from this site wich you dont even bother to give a link, thoses scores are at 15W TDP according to AMD s slides but for some reason this shill site did blatantly lie and did put 35W on their viral pics.

Other than the fact that i7-7500W (15W) is still considerably faster than their 35W offering - how can FX-9800P beat the performance of a higher-power part?

35W according to a site that lie, indeed that s the kind of infos you need to make a point.

Other than that if 15W were enough for what is an i3 to score 4.03 in CB 11.5 then a i7 at same frequency would score 8.06 at barely 30W, wich is not the case but you believe in such impossiblities nevermind the numbers.

You can look the things the other way around, a i7 6700K use 75W in CB R15 for a 885 score, at half the cores count it would score 442 at about 40W and to score the alleged 363 at "15W" those 2C/4T would use 40.(363^2)/(442^2) = 27W, wich is about the PL1 of the "15W" KBL..

Laws of physics are straightfoward but certainly that they are not much compatible with marketing fairy tales..

 

[Sweepr]

For one that s not the AMD presentation but viral marketing from this site wich you dont even bother to give a link, thoses scores are at 15W TDP according to AMD s slides but for some reason this shill site did blatantly lie and did put 35W on their viral pics.

False. Tech Report confirms it was at 35W according to AMD. Does this mean FX-9800P results are exceeding 15W?

AMD says a 35W Bristol Ridge APU turned in a Cinebench R15 single-threaded score of 93.24 in its press materials.

http://techreport.com/news/29942/amd-lifts-the-curtain-a-little-bit-on-its-bristol-ridge-apus

You can look the things the other way around, a i7 6700K use 75W in CB R15 for a 885 score, at half the cores count it would score 442 at about 40W and to score the alleged 363 at "15W" those 2C/4T would use 40.(363^2)/(442^2) = 27W, wich is about the PL1 of the "15W" KBL..

Laws of physics are straightfoward but certainly that they are not much compatible with marketing fairy tales..

So basically you have zero proof of what you're saying, and you're trying to draw conclusions about 15W Kaby Lake-U from 95W Skylake-S.

Since you like repetitions, let's post this image till you admit that you're either wrong or making stuff up by saying i7-7500U used 25-27W during CB R15 MT:

 

[mikk]

The Kabylake system was clearly configured as 15W as I can see from the average which is locked down at 15W. Burst power can be higher for a short period. Only some deluded is going to disagree.

 

[Abwx]

False. Tech Report confirms it was at 35W according to AMD. Does this mean FX-9800P results are exceeding 15W?

http://techreport.com/news/29942/amd-lifts-the-curtain-a-little-bit-on-its-bristol-ridge-apus

Lol, like you they are just parroting what the shill site published and they got the "info" from there apparently :

http://wccftech.com/amd-bristol-ridge-apu-computex-launch/

What s written on note 6..?..

So whatever lie is good for you, provided it downplay AMD.

So basically you have zero proof of what you're saying, and you're trying to draw conclusions about 15W Kaby Lake-U from 95W Skylake-S.

I showed you above who is the one who is accurate when talking technology and numbers particularly, do you need more proof that you are not interested in tech other than from a brand loyalty marketing POV..?.

Other proof is that i have a background in electronics while you are stuck thinking that physics law are irrelevant when it comes to Intel, wich is true in respect of Intel s marketing, of course.

Since you like repetitions, let's post this image till you admit that you're either wrong or making stuff up by saying i7-7500U used 25-27W during CB R15 MT:

The only thing that can be admitted is that they polished the numbers, when hot and in HEVC the laptop get the cores (without the rest of the SoC...) to 16.7W and yet they pretend to measure less than this from the main when benching, that s ridiculous and is just aknowledgement that the MB drain power simultaneously from the main and the battery, hence the former being lower than what HVinfo can display..

See..?.. At 16.7W for the cores the CPU is not at 100% but at about 80%, wich mean that when the laptop is hot 20W is the actual TDP for the cores only, so how much when it s cold and the bench launched.?

FTR the 25W PL1 is not a limitation to 25W but the maximum over a given period, wich means that it can go higher..

 

[Sweepr]

So whatever lie is good for you, provided it downplay AMD.

The results speak for themselves. Kaby Lake just extends Intel's lead on mobile. Considering that Bristol Ridge scores 93 pts at 35W according to the slide you just posted (you can apologize to Hardware Canucks later) - we can easily estimate the MT scores based on AnandTech's Excavator testing: 35W BR should score ~322 pts @ Cinebench R15 MT

This means an i7-7500U is faster (363 pts vs ~322 pts) at a fraction of the TDP (15W vs 35W) - and Bristol Ridge should get trounced by 28W models and future 4C/8T 'U' parts.

The only thing that can be admitted is that they polished the numbers, when hot and in HEVC the laptop get the cores (without the rest of the SoC...) to 16.7W and yet they pretend to measure less than this from the main when benching, that s ridiculous and is is just aknowledgement that the MB drain power simultaneously from the main and the battery, hence the former being lower than what HVinfo can display..

No, it shows 14.9W-18.2W during the entire CB R15 MT run + 14.898W CPU package power, which fits the TDP rating of this chip. What you said above (keeping short term burst 25W PL2 for >2 minutes) is a lie according to NBC itself, and after a dozen of posts you still have zero data to back it up.

 

[Abwx]

The results speak for themselves. Kaby Lake just extends Intel's lead on mobile.

I would say lead in marketing..

Considering that Bristol Ridge scores 93 pts at 35W according to the slide you just posted (you can apologize to Hardware Canucks later)

Why should i apologise to liars when the slides say explicItly that the 232.86 score is for a FX@15W...?

As for the TDP and ST score there s nothing of the sort written on the slide, why are you keeping making up numbers that do not exist and then have the guts to state that they are from AMD..?.

No, it shows 14.9W-18.2W during the entire CB R15 MT run + 14.898W CPU package power, which fits the TDP rating of this chip. What you said above (keeping short term burst 25W PL2 for >2 minutes) is a lie according to NBC itself, and after a dozen of posts you still have zero data to back it up.

They measure the power at the main, so thoses numbers are impossible even if the CPU was stuck at 15W, there would be the losses in two DC conversion plus the RAM, screen and so on, but since you believe in such miracles, why not, but just dont pretend that others should be that gullible, and as said, it contradict the most elementary laws of physics, if thoses numbers were true then a i7 6700K would score 12.74 in CB R11.5...

 

[Sweepr]

why are you keeping making up numbers that do not exist and then have the guts to state that they are from AMD..?.

I'm sorry, what did you say again?

http://www.guru3d.com/news-story/amd-to-launch-7th-gen-mobile-apus-at-computex.html
https://tweakers.net/nieuws/109983/amd-kondigt-zevende-generatie-mobiele-apus-op-computex-aan.html
http://www.extremetech.com/computin...s-new-bristol-ridge-plans-for-computex-launch

Now just multiply it by the usual Excavator 2M/4C scaling and you can see that they are still miles behind in CPU performance-per-watt. Kaby Lake's real competition are current Skylake designs.

They measure the power at the main, so thoses numbers are impossible even if the CPU was stuck at 15W, there would be the losses in two DC conversion plus the RAM, screen and so on, but since you believe in such miracles, why not, but just dont pretend that others should be that gullible, and as said, it contradict the most elementary laws of physics, if thoses numbers were true then a i7 6700K would score 12.74 in CB R11.5...

I understand that you have zero proof to back up your claims, but 95W Skylake-S results don't count as proof that a mobile processor on an improved process (i7-7500U) draws >65-80% more than what a very well known / respected website who reviewed the chip published. Neither do your mirabolant theories about NBC's methodology. Kaby Lake-U didn't sustain >25W (PL2 short bursts) for over 2 minutes during CPU-only benchmarks, that's a lie.

As mentioned above, the Core i7-7500U cannot maintain its maximum clock with a TDP limit of 15 Watts, but 3.3-3.2 GHz are still much higher than a comparable Skylake model, which would hit the 15-Watt limit at 2.8 GHz.

 

[LTC8K6]

Why is AMD not screaming about the wattage cheating, Abwx? Surely they can see what you can see, and would at least point it out?

 

[Abwx]

Why is AMD not screaming about the wattage cheating, Abwx? Surely they can see what you can see, and would at least point it out?

Why should they bother with shill sites that publish viral marketing..?.

I posted the slide above that state that CB R15 score of 232.86 is for a FX at 15W, it is below on note 6 and if those sites didnt correct their viral graph it just mean that it s willfull viral marketing sites that do not deserve to be adressed by AMD.

Now if at AMD there s someone that look at the e-press he should make sure that those sites never get any CPU or GPU, that s how they should adress Canuckistan and other such shills and that s what i would call a good PR management, just like they apparently did with Notebookcheck...

That being said thanks to Sweepr, althoug unwillfully, we know wich are the biaised and unworthy sites..

Notice that Sweepr reference to a 93 ST score at allegedly 35W, wich he obviously extracted from those sites viral graphs and wich is not mentionned anywhere by AMD, not only it doesnt exist but show that the viral sites are cluless since in ST Bristol Ridge use only one core while at full throughput it use 4, so in ST the power budget for a single core is such that they could almost double the base frequency they have at 15W, that get us at the 3.6GHz at 15W and right to the 93 score the shill site are talking about..

Btw, not so much with 15W dual cores since in ST this require half the power budget just to be at the 15W MT stock frequency..

 

[Sweepr]

Notice that Sweepr reference to a 93 ST score at allegedly 35W, wich he obviously extracted from those sites viral graphs and wich is not mentionned anywhere by AMD

False.

Edit: Found the official AMD presentation (slide 24):

Cinebench R15 1T is used to represent per core
performance; the 2016 7th Generation AMD FX @ 35W scored 93.24 while the 6th Generation AMD FX @ 35W scored 81.82

http://phx.corporate-ir.net/External.File?item=UGFyZW50SUQ9MjU3MzM4fENoaWxkSUQ9LTF8VHlwZT0z&t=1

AMD says a 35W Bristol Ridge APU turned in a Cinebench R15 single-threaded score of 93.24 in its press materials.

http://techreport.com/news/29942/amd-lifts-the-curtain-a-little-bit-on-its-bristol-ridge-apus

AMD is also talking up its Cinebench scores, as shown below, but the company’s reported baseline benchmarks are a bit off compared to what we’ve seen in retail. We elected to create our own graph for this data since the slide isn’t completely clear.

http://www.extremetech.com/computin...s-new-bristol-ridge-plans-for-computex-launch

http://www.guru3d.com/news-story/amd-to-launch-7th-gen-mobile-apus-at-computex.html
https://tweakers.net/nieuws/109983/amd-kondigt-zevende-generatie-mobiele-apus-op-computex-aan.html
http://bbs.kakaku.com/bbs/-/SortID=19761921/

Have to admit that your adjectives are funny though - 'viral marketing websites', 'shill site', 'biased and unworthy', 'clueless'